Journal list menu

Volume 592, Issue 11 p. 1837-1846
Research Letter
Free Access

A common polymorphic variant of UGT1A5 displays increased activity due to optimized cofactor binding

Fan Yang

Fan Yang

School of Pharmaceutical Science and Technology, Health Sciences Platform, Tianjin University, China

Search for more papers by this author
David Machalz

David Machalz

Pharmaceutical and Medicinal Chemistry, Computer-Aided Drug Design, Institute of Pharmacy, Free University Berlin, Germany

Search for more papers by this author
Sisi Wang

Sisi Wang

School of Pharmaceutical Science and Technology, Health Sciences Platform, Tianjin University, China

Search for more papers by this author
Zhengyi Li

Zhengyi Li

School of Pharmaceutical Science and Technology, Health Sciences Platform, Tianjin University, China

Search for more papers by this author
Gerhard Wolber

Gerhard Wolber

Pharmaceutical and Medicinal Chemistry, Computer-Aided Drug Design, Institute of Pharmacy, Free University Berlin, Germany

Search for more papers by this author
Matthias Bureik

Corresponding Author

Matthias Bureik

School of Pharmaceutical Science and Technology, Health Sciences Platform, Tianjin University, China

Correspondence

M. Bureik, School of Pharmaceutical Science and Technology, Health Sciences Platform, Tianjin University, 92 Weijin Lu, Tianjin 300072, China

Fax: +86- 22-874018304

Tel: +86- 22-87401830

E-mail: [email protected]

Search for more papers by this author
First published: 30 April 2018
Citations: 10
Edited by Judit Ovádi

Abstract

Uridine diphosphate-glucuronosyltransferases (UGTs) are the most important phase II enzymes in human drug metabolism. Using permeabilized recombinant fission yeast cells (enzyme bags), we demonstrate that UGT1A5 can catalyze an N-glucuronidation reaction. We characterized two new polymorphic UGT1A5 variants: a common ninefold mutant (UGT1A5*8) with double-fold activity and a much rarer sixfold mutant (UGT1A5*9), which has the same activity as the wild-type. Molecular modeling studies indicate that the minor effects of all mutations, except for Gly259Arg, are due to their distance to the substrate binding site. Extensive molecular dynamics simulations revealed that the Gly259Arg mutation stabilizes helix Q through a newly formed hydrogen bonding network, which places the cofactor in a much more favorable geometry in UGT1A5*8 as compared to the wild-type.

Abbreviations

CYP, cytochrome P450

MD, molecular dynamics

MIST, metabolites in safety testing

RMSD, root-mean-square deviation

RMSF, root-mean-square fluctuation

UDP-GA, uridine diphosphoglucuronic acid

UGT, uridine diphosphate-glucuronosyltransferase

Uridine diphosphate-glucuronosyltransferases (UGTs) are a superfamily of conjugation enzymes that use the cofactor uridine diphosphate-glucuronic acid (UDP-GA) to catalyze the glucuronidation of numerous substrates at a variety of functional groups [1]. Glucuronidation of endogenous substrates plays an important role in the regulation of endobiotics levels and the maintenance of homeostasis while UGT-dependent bioconversion of xenobiotics is considered to be the most important phase II reaction type in human drug metabolism [2]. There are 22 human UGT enzymes which belong to four different families (UGT1, UGT2, UGT3, and UGT8); out of these, the 19 members of the UGT1 and UGT2 families are currently thought to be significantly involved in drug metabolism. Like other drug metabolizing enzymes (such as cytochrome P450s), some UGTs are highly polymorphic in humans and are affected by common inherited polymorphisms and copy-number variations, in addition to rare genetic alterations at the origin of single-gene disorders [2]. Such variations are present at divergent frequencies among different ethnic populations and are thought to contribute to variable drug clearance and response [X. Cao, P. Durairaj, Y. Fan & M. Bureik, in revision]. Previously, we have developed a recombinant UGT expression system based on the fission yeast Schizosaccharomyces pombe and used the resulting strains for glucuronide production in whole-cell biotransformations [3-6]. However, in such an assay test compounds have to pass through several biological barriers (cell wall, cell plasma membrane, endoplasmic reticulum membrane) before reaching the UGT enzymes; similarly, UGT reaction products have to pass through the same barriers before they can be detected in and purified from the culture supernatant. For the glucuronic products and some polar substrates, this step may well be the main obstacle for higher biotransformation efficiencies, and consequently, the UGT-dependent metabolism of compounds that do not migrate readily through these barriers is difficult or even impossible to analyze. This is especially true for N-glucuronidation reactions, as many amines are protonated at physiological pH values. To overcome these limitations, we have recently developed a new approach that makes use of permeabilized fission yeast cells [7]. In this method, recombinant fission yeast cells are permeabilized with Triton X-100 prior to the biotransformation reaction. The resulting permeabilized cells (‘enzyme bags’) retain most cellular structures, protein-protein interactions, and most of their intracellular enzymes in their original environment [8, 9]. We demonstrated that this method yields porous cells which let small molecules diffuse free between the buffer system and the inside of the cells; at the same time, larger molecules such as enzymes are kept within the cells. In this study, we transfered this method to fission yeast cells that recombinantly express human UGTs and demonstrated its use for monitoring an N-glucuronidation reaction. N-glucuronidation is an important pathway for metabolism and disposition of many amines, and within the context of metabolites in safety testing (MIST), N-glucuronides are also of interest because they tend to be formed much more in humans than in many other animal species and as such often end up as disproportionate metabolites [10]. UGT1A4 and UGT2B10 are currently being thought to be the most important enzymes catalyzing this reaction type in humans, while UGT1A3 is contributing only to a smaller extent [11]. By contrast, the involvement of UGT1A5, which displays a very high sequence homology to both UGT1A3 and UGT1A4, in N-glucuronidation is unknown. UGT1A5 expression levels are low in liver, kidney and the gastrointestinal tract [12]; it was reported to strongly metabolize 1-hydroxypyrene, but shows only low rates of 4-methylumbelliferone and scopoletin glucuronidation and, surprisingly, does not glucuronidate 4-aminobiphenyl, which is a good substrate for both UGT1A4 and UGT1A3 [13]. Thus, no N-glucuronidation reaction has yet been reported for UGT1A5, even though is close relationship with UGT1A3 and UGT1A4 suggests that it can catalyze this reaction type.

Materials and methods

Materials

UDP-glucuronic acid and UDP-glucose were from Sigma-Aldrich (St. Louis, MO); ammonium hydrogen carbonate and potassium chloride were from Jiangtian Chemical (Tianjin, China); Tris-HCl was from AKZ-Biotech (Tianjin, China); Triton X-100 (1%) was from Leagene (Beijing, China); glycerol was from Dingguo (Tianjin, China); UGT-Glo™ Assay V2082 was from Promega (Madison, USA); white 96-well microtiter plates were from Nunc (Thermofisher scientific, Lagenselbold, Germany); human liver microsomes (HLMs) were from Sekisui XenoTech (Kansas City, KS, USA); DMSO for molecular biology was from Sigma-Aldrich (St. Louis, MO). All other chemicals and reagents used were of the highest grade available.

Fission yeast strains and media

Except for YF1 and YF2 (see below), all fission yeast strains used in this study have been described previously [3, 5, 6]; they are listed in Table 1. The preparation of media and basic manipulation methods of S. pombe were carried out as described [14]. Strains were generally cultivated at 30 °C in Edinburgh Minimal Medium (EMM) with supplements of 0.1 g·L−1 final concentration as required. Liquid cultures were kept shaking at 150 r.p.m. Construction of fission yeast strains YF1 and YF2 was done following the strategy described previously [6]. Briefly, the DNAs coding for human UGT1A5*8 and UGT1A5*9 were synthesized by General Biosystem (Anhui, China) and cloned via Nde I and BamH I into the integrative vector pCAD1 [15]. In pCAD1, recombinant expression is controlled by the strong endogenous nmt1 promotor [16]. Transformed cells were grown on EMM dishes with 5 μm thiamine to allow better growth under repressed conditions. The newly created pCAD1-UGT plasmids were used to transform the parental S. pombe strain NCYC2036 [17] as described previously [18] to yield strains YF1 and YF2.

Table 1. Fission yeast strains used in this study
Strain Expressed protein Parental strain Genotype References
CAD200 UGT1A9 NCYC2036 h¯ura4-D18 leu1::pCAD1-UGT1A9 [6]
DB2 UGT1A4 NCYC2036 h¯ura4-D18 leu1::pCAD1-UGT1A4 [3]
DB4 UGT2B7*2 NCYC2036 h¯ura4-D18 leu1::pCAD1-UGT2B7*2 [5]
DB7 UGT2B7*1 NCYC2036 h¯ura4-D18 leu1::pCAD1-UGT2B7*1 [5]
DB22 UGT1A5*1 NCYC2036 h¯ura4-D18 leu1::pCAD1-UGT1A5 [3]
DB31 UGT2B11 NCYC2036 h¯ura4-D18 leu1::pCAD1-UGT2B11 [3]
DB34 UGT2B28 NCYC2036 h¯ura4-D18 leu1::pCAD1-UGT2B28 [3]
YF1 UGT1A5*8 NCYC2036 h¯ura4-D18 leu1::pCAD1-UGT1A5*8 This study
YF2 UGT1A5*9 NCYC2036 h¯ura4-D18 leu1::pCAD1-UGT1A5*9 This study

Biotransformation with enzyme bags

Fission yeast strains were grown in 10 mL liquid culture of EMM with supplements as needed at 30 °C and 230 r.p.m. for 24 h. For each assay, 5 × 107 cells were transferred to 1.5 mL Eppendorf tubes, pelleted and incubated in 1 mL of 0.3% Triton-X100 in Tris-KCl buffer (200 mm KCl, 100 mm Tris-Cl pH 7. 8) and incubated at room temperature for 60 min at 150 r.p.m. for permeabilization. Cells were then washed thrice with 1 mL NH4HCO3 buffer (50 mm, pH 7.8) and after the final wash the cell pellet was either frozen or directly used for UGT-dependent reactions. For storage, 200 μL Tris-Cl pH 7.8 with 0.1% glycerol were added and samples stored at −80 °C. For UGT-Glo assays, 20 μL reaction system and 10 μL UDPGA (16 mm) were added and samples were incubated for 3 hrs (37 °C, shaking at 1000 r.p.m.). The final substrate concentration was 20 μm for the UGT multi-enzyme substrate (UGT-glo substrate A) and 60 μm for the UGT1A4 substrate (UGT-glo substrate B). Afterwards, the reaction mixture was transferred into 1.5 mL Eppendorf tubes and centrifuged at 16 000 g for 1 min. The supernatant was then transferred to white 96-well microtiter plates for luminescence measurements.

Bioluminescence detection

Reaction supernatants were transferred to the white microtiter plates and an equal amount of reconstituted luciferin detection reagent was added to each well. Plates were then incubated at room temperature for 20 min and luminescence was recorded on a Magellan infinite 200Pro microplate reader (Tecan; Männedorf, Switzerland). All measurements were done at least three times in triplicates.

Statistical analysis

All data are presented as mean ± standard deviation. Statistical significance was tested using two-tailed T-test. Differences were considered significant if < 0.05. Statistical analysis was done using GraphPad Prism 5. 01 (GraphPad Software, Inc., La Jolla, CA, USA).

Analysis of genetic variants from exome sequences and bioinformatic filtering

ExAC MAF data were collected from the Exome Aggregation Consortium (ExAC) release GRCh38, which includes exome data from 60 706 individuals of diverse ethnicities [19]. From these data, we extracted the single nucleotide polymorphisms (SNPs) in the UGT1A5 gene that lead to missense variants and have a minor allele frequency (MAF) ≥ 0.01. For this purpose, only the UGT1A5 reference standard sequence (NCBI Reference Sequence: NP_061951.1) contained in the RefSeqGene project [20] was used, transcript variants were not included. For the nine SNPs identified in this way we also extracted the allele frequency data for five major populations (East Asians, Europeans, Africans, admixed Americans, and South Asians, respectively) from the 1000 Genomes Project [21].

Homology modeling

Homology modeling was performed on the SWISS MODEL server [22] for the wild-type and the ninefold mutant (UGT1A5*8) of UGT1A5. Sequence homology search yielded UDP-glucuronosyl/UDP-glucosyltransferase as best-ranked template (PDB entry 2PQ6, 18.45% sequence identity, 77% coverage). Since this crystal structure does not include a co-crystallized cofactor and sequence alignment does not match the catalytic region of UGT1A5, sterol 3-beta-glucosyltransferase (PDB code 5GL5) was chosen as template for the homology model. The coordinates for the cofactor UDP-glucose were taken as co-crystallized in PDB entry 5GL5 and changed into UDP glucuronic acid (UDP-GA) using the MOE software package (Molecular Operating Environment 2015. 10; Chemical Computing Group ULC, Montreal, Canada). Side chains in the cofactor binding site were rotated to allow for accommodation of the cofactor.

Molecular dynamics simulations

Extensive all-atom Molecular Dynamics (MD) simulations were performed for 100 ns in three replicates for the wild-type and the ninefold mutant of UGT1A5 in complex with the cofactor UDP-GA. The systems were solvated using TIP3P water [23] in an orthorhombic box in Maestro (Schrödinger Release 2017-3: Maestro, Schrödinger, LLC, New York, NY, 2017). Chloride atoms were added to counter the positive charge of the systems and additional 0.15 m sodium chloride was added to simulate physiological conditions. Different seeds were used to randomize starting velocities. The simulations of the 50–52 k atom systems were carried out in the GPU-accelerated mode of Desmond [24] (version 5.1) using the OPLS-AA force field [25] on a water-cooled NVidia GTX1070 graphics card. During the simulations, the RESPA integrator calculated the intermolecular forces with a distance cut-off of 9 Å for electrostatic interactions. The temperature of 300 K was kept constant with the Nose-Hoover chain method. Constant pressure of 1.01325 bar was maintained with the Martyna-Tobias-Klein Barostad method. Atom coordinates were written out every 20 ps. Root-mean-square deviation, (RMSD), root-mean-square fluctuation (RMSF), and hydrogen bonding analysis was carried out in Maestro. RMSD was calculated based on all heavy atoms and the RMSF on Cα atoms. Default hydrogen bonding geometry criteria in Maestro were kept. The Cα distance of residue 259 and 401 was measured in VMD [26]. Further data analysis and plotting was performed using the Python programming language.

Docking

Molecular docking was conducted for UGT-Glo substrates A and B using GOLD (version 5.2) [27] with 30 genetic algorithm runs. A representative from the second replica of UGT1A5*8 simulation served as protein conformation. The search efficiency was set to 200% while keeping default settings for all other parameters. The binding sphere radius was set to 10 Å with the anomeric C atom of the glucuronosyl moiety as center. The obtained docking poses were analyzed in LigandScout (version 4.09) [28-30] after energy minimization using the MMFF94 force field [31].

Results and Discussion

Identification of new UGT probe reactions

Recently we reported the preparation of permeabilized cells (enzyme bags) from fission yeast cells that recombinantly express human UDP-glucose 6-dehydrogenase or human cytochrome P450 enzymes by permeabilization with 0.3% (v/v) Triton X-100 and their use for efficient biotransformation [7, 32]. In this study, we wanted to transfer this method to fission yeast strains that recombinantly express members of the human UGT1 and UGT2 families. For this purpose, enzyme bags were prepared from these strains and their activity was monitored using the UGT multienzyme probe substrate 6-hydroxy-4-methyl-1,3-benzothiazole-2-carbonitrile (UGT-Glo substrate A). In this assay, conjugation of the luminogenic substrate (prior to conversion into a luciferin derivative using D-cysteine) correlates to a decrease in luminescence signal when incubated with luciferase. UGT-Glo substrate A is a good substrate for both UGT1A9 and UGT2B7 but much less so for UGT1A4 (according to the instructions of the manufacturer). In addition to these enzymes, we also tested UGT1A5, UGT2B11, and UGT2B28, which are of interest for a variety of reasons: As mentioned in the introduction, the question of whether UGT1A5 can catalyze N-glucuronidation still remained to be answered. UGT2B28 is near-identical to its neighbor gene UGT2B11 and it is one of the most commonly deleted genes in humans [33]. However, while UGT2B11 was reported to glucuronodate 12-hydroxyeicosatetraenoic acid, 15-hydroxyeicosatetraenoic acid, and 13-hydroxyoctadecadienoic acid, respectively [34], UGT2B28 metabolizes several steroids and xenobiotics [35]; it is considered to be the main androgen-inactivating UGT isoform under high-androgen exposure [36]. It was found that the UGT-Glo substrate A could be metabolized by all UGTs tested (Fig. 1). As expected, activity was high using either UGT1A9 (78 ± 12 μm·day−1) or UGT2B7, with the activities of the two most common polymorphic variants UGT2B7*1 (62 ± 7 μm·day−1) and UGT2B7*2 (56 ± 10 μm·day−1) being not significantly different from each other. Interestingly, UGT2B28 (54 ± 6 μm·day−1) displayed a comparable activity level as well. By contrast, UGT2B11 had the lowest activity of all enzymes tested (11 ± 8 μm·day−1), but its activity was still significantly higher than the background. Activity of UGT1A5 (36 ± 9 μm·day−1) was not significantly different to that of UGT1A4 (32 ± 10 μm·day−1). Taken together, these new probe reactions demonstrate the usefulness of enzyme bags for recombinant UGT activity assays and, moreover, permit convenient activity screening of UGT1A5, UGT2B11, and UGT2B28.

Details are in the caption following the image
Activity of seven UGT isoenzymes toward UGT-Glo substrate A. Enzyme bags were prepared from seven strains (recombinantly expressing human UGT1A9, UGT1A4, UGT1A5, UGT2B7*1, UGT2B7*2, UGT2B11, or UGT2B28, respectively) as indicated and activity was monitored using UGT-Glo substrate A. Data shown are given as percent substrate consumed after 3 h (substrate conc. 20 μm).

Bioinformatic analysis of UGT1A5 missense variants and activity comparison of different UGT1A5 haplotypes

A recent study done by the Exome Aggregation Consortium (ExAC; http://exac. broadinstitute. org) describes the joint variant calling and analysis of high-quality variant calls across more than 60 000 human exomes [19], which considerably exceeds previously available exome-wide variant databases such as that of the 1000 Genomes Project [21]. In a recent study, we utilized this data set to provide a substantially increased resolution for the analysis of genetic variants that lead to missense mutations in CYPs and UGTs [X. Cao, P. Durairaj, Y. Fan & M. Bureik, in revision]. In the course of that investigation it was found that the pattern of missense single nucleotide polymorphisms (SNPs) in UGT1A5 is very unique among all UGTs: There are six SNPs with an average minor allele frequency (MAF) of 0.1208 ± 0.0002 and three others with an average MAF of 0.1157 ± 0.0006 (Table 2). Eight of these nine variations (all except His225Tyr) where previously found in the 1000 Genomes Project with higher frequencies in people from east Asia (average MAF 0.23) and south Asia (average MAF 0.22) and lower frequencies in Europeans (average MAF 0.10), Americans of mixed ancestry (average MAF 0.10), and Africans (average MAF 0.08). Three of these SNPs (Leu63Pro, Ala158Gly, and Gly259Arg) were also previously found in a study on genetic variations in the Chinese population, albeit at somewhat lower frequencies (average MAF 0.19) than those reported in the 1000 Genomes Project for Asians [37]. From these data, we predict that there are two common haplotypes: UGT1A5*1 (wild-type) with a frequency of > 0.8 and secondly a ninefold mutant (Leu48Ser, Asp50Glu, Leu63Pro, His142Asn, Ala158Gly, His225Tyr, Leu248Ile, Val249Leu, Gly259Arg) with a frequency of 0.115. In addition, there should also be a rarer haplotype of the sixfold mutant (Leu48Ser, Asp50Glu, Leu63Pro, His142Asn, Ala158Gly, His225Tyr) with a frequency of 0.005. As of February 2018, the UGT Alleles Nomenclature Home Page [38] lists seven UGT1A5 haplotypes. We therefore suggest to name the ninefold mutant UGT1A5*8 and the sixfold mutant UGT1A5*9. To judge the functionality of these two mutants, the DNAs coding for them were cloned into the integrative fission yeast vector pCAD1 [15] and the newly created pCAD1-UGT plasmids were used to transform the S. pombe strain NCYC2036 [17] to yield strains YF1 and YF2, respectively (Table 1). Enzyme bags prepared from these two strains as well as from strain DB22 (recombinantly expressing human UGT1A5*1) were used for the glucuronidation of UGT-Glo substrates A and B (UGT1A4 substrate) to monitor O- and N-glucuronidation activity, respectively (Fig. 2). Surprisingly, it was found that for either substrate the sixfold mutant UGT1A5*9 (28 ± 18 μm·day−1 for UGT-Glo substrate A and 104 ± 52 μm·day−1 for UGT-Glo substrate B) has the same activity as the wild-type (35 ± 12 μm·day−1 and 121 ± 37 μm·day−1), while the activity of the ninefold mutant UGT1A5*8 is twice as high in both cases (70 ± 4 μm·day−1 and 256 ± 78 μm·day−1). Thus, both polymorphic variants retain their full functionality, which in view of the considerable number of mutations in comparison to the wild-type is a remarkable finding. Moreover, for the probe reactions used here UGT1A5*8 is an activating polymorphic missense variant, which is a very rare occurence among either UGTs or human drug metabolizing enzymes in general, as most of their missense variants tend to reduce or abolish enzyme activity [39]. Finally, these data demonstrate for the first time that UGT1A5 can efficiently catalyze a N-glucuronidation reaction.

Table 2. Previously published sequence variations in the UGT1A5 gene. Shown are the nine most common SNPs that lead to missense mutations. The first six variations are shared by UGT1A5*8 and UGT1A5*9, while the last three are only present in UGT1A5*8. The SNP that causes the Gly259Arg mutation is shown in bold. SNP, single nucleotide polymorphism; ExAC, exome aggregation consortium; CHH, Chinese from Henan; EAS, East Asians; EUR, Europeans; AFR, Africans; AMR, Admixed Americans; SAS, South Asians; n. d., not detected
Reference SNP Allele frequency
ExACa CHHb EASc EURc AFRc AMRc SASc
rs3755323 0.1209 0.187 0.2331 0.0964 0.0809 0.1009 0.2219
rs3755322 0.1210 0.187 0.2331 0.0964 0.0809 0.1009 0.2219
rs3755321 0.1209 0.191 0.2331 0.0964 0.0809 0.1009 0.2219
rs3755320 0.1206 0.189 0.2331 0.0964 0.0809 0.1009 0.2188
rs12475068 0.1209 0.189 0.2331 0.0964 0.0809 0.1009 0.2219
rs17862867 0.1205 0.189 n. d. n. d. n. d. n. d. n. d.
rs2012736 0.1160 0.189 0.2331 0.0964 0.0809 0.1009 0.2178
rs17862868 0.1160 0.183 0.2331 0.0964 0.0809 0.1009 0.2178
rs3892170 0.1149 0.183 0.2331 0.0964 0.0809 0.1009 0.2178
  • a Data from [19].
  • b Data from [37].
  • c Data from [21].
Details are in the caption following the image
Influence of polymorphic variants on the activity of UGT1A5. Enzyme bags prepared from three strains (recombinantly expressing human UGT1A5*1, UGT1A5*8, or UGT1A5*9, respectively) as indicated were incubated either with UGT-Glo substrate A (black columns) or with UGT-Glo substrate B (gray columns). Data shown are given as percent substrate consumed after 3 h (substrate conc. 60 μm). **< 0.01, ****< 0.001 vs. UGT1A5*1.

Homology modeling for UGT1A5

To investigate the molecular nature of the increased enzymatic activity of the UGT1A5*8 mutant for both UGT-Glo substrates we built structural models of the wild-type UGT1A5*1 and the mutant UGT1A5*8. Due to the lack of x-ray structures, homology models were built from the crystal structure of sterol 3-beta-glucosyltransferase (PDB 5GL5) as shown in Fig. 3. While the C-terminal domain is responsible for cofactor binding, the N-terminal domain is responsible for substrate recognition [40]. Since the crystal structure of the C-terminal domain of human UGT2B7 is available, its geometric deviation can serve as control for the quality of our homology model (Fig. 3B) showing a root-mean-square deviation of Cα carbon coordinates of 2.1 Å. Hence, the high structural similarity with the only available human x-ray structure of UGT2B7 C-term domain shows the validity of residues 285 to 446 of the homology models of UGT1A5*1 and UGT1A5*8. The high quality of both homology models is further underlined by the absence of severe phi-psi angle outliers (Fig. S3). The catalytic dyad of His40 and Asn197 as well as His373 show a correct orientation toward the cofactor uridine 5′-diphosphoglucoronic acid (UDP-GA) in both models in accordance to the previously stated catalytic site geometry [40].

Details are in the caption following the image
(A) Structural homology model for UGT1A5*8 with bound cofactor Uridine-diphosphoglucoronic acid (UDP-GA). The secondary structure ribbon is shown in gray. Helix Q is highlighted in blue. The cofactor (colored turquois) is situated in the catalytic cleft between the N-terminal (left) and C-terminal (right) domains. (B) Superposition of C-terminal domains of the UGT2B7 crystal structure (yellow) and UGT1A5*8 homology model (gray) with a root-mean square root of 2. 1 Å. (C) Cofactor protein interaction diagram for UGT1A5 and UDP-GA. The glucuronic acid moiety of UDP-GA is hold in place by electrostatic interaction with Arg174 and hydrogen bonding to Ser376 and Asp397. The uridine-diphosphate moiety forms hydrogen bonds to Ser307, Leu308, His373, His377, and Gly378. Blue double-headed arrow represents electrostatic interaction. Red arrows represent hydrogen bond acceptance and green arrow hydrogen bond donation.

Comparative molecular dynamics simulations of UGT1A5*1 and UGT1A5*8

When looking for the mutations responsible for the increase in catalytic activity we limited our view on the three mutations (Leu248Ile, Val249Leu and Gly259Arg) present only in the ninefold but not in the sixfold mutant. The changes of Leu248 to Ile and Val249 to Leu are unlikely to show a large effect on the function because these substitutions are conservative and, in addition, they are not located near the catalytic site. Hence, the Gly259Arg mutation alone seems to be responsible for the increased catalytic activity. Gly259 is not directly involved in cofactor or substrate binding (data not shown). However, we observe hydrogen bonds from the mutated Arg259 to Asp400 and Asn401 located in helix Q (Fig. 4) in the ninefold mutant UGT1A5*8. Since we were interested in the impact of the Gly259Arg mutation on the stability of the hydrogen bonds proximal to the cofactor binding site, we conducted molecular dynamics (MD) simulations. These are a well-established method [41] allowing insight into the structural dynamics of an enzyme influencing its function. The homology models of wild-type UGT1A5*1 and ninefold mutant UGT1A5*8 holoenzymes were simulated in three replicates for 100 ns in Desmond [25]. Both UGT1A5*1 and UGT1A5*8 remain stable during 100 ns of MD simulation (Fig. S1B). Simulation results of UGT1A5*8 show a constantly lower distance between the Cα atoms of Arg259 and Asn400, both situated in the helix Q, compared to wild-type UGT1A5*1 (Fig. S1A). The constant Cα distance of 11-12 Å is caused by hydrogen bonding of Arg259 to Asp400 and Asn401, which is not present for Gly259 in UGT1A5*1. Hence, helix Q is allowed to move more freely in UGT1A5*1 than in UGT1A5*8 with the additional hydrogen bonding of Arg259. Stabilization of helix Q in UGT1A5*8 is apparent from lower root-mean-square fluctuation (RMSF) of the Cα atoms compared to UGT1A5*1. Helix Q also comprises Asp397 and Gln398 contributing to cofactor binding via hydrogen bonding to the 3′-hydoxyl and 4′-hydroxyl groups of UDP-GA (Fig. 4). UGT1A5*8 shows more frequent hydrogen bonding to UDP-GA via these two residues than UGT1A5*1 throughout the three simulations (Fig. S2A). To describe the interaction between helix Q and the cofactor, hydrogen bond occurrence between Asn397/Gln398 and the cofactor UDP-GA was investigated (hydrogen occupancy, Fig. S2B). The lower hydrogen bond occupancy in the first simulation replicate of UGT1A5*1 occurs due to partial unbinding of the glucuronic acid moiety of UDP-GA terminating the hydrogen bonding to Asn397/Gln398. Fig. S2A shows the number of hydrogen bonds between Asn397 and Gln398 and UDP-GA: UGT1A5*8 simulations did not only show higher hydrogen bonding occupancy, but also a higher overall frequency of hydrogen bonds between UDP-GA and helix Q.

Details are in the caption following the image
Hydrogen bond network ranging from Arg259 to cofactor UDP-GA in UGT1A5*8. Arg259 forms hydrogen bonds to Asp400 and Asn401 in the helix Q. Asp400 shows hydrogen bonding to Asp397 with the helix. Asp397 and Gln398 form hydrogen bonds to the cofactor UDP-GA (turquois stick representation).

Substrate docking experiments for UGT1A5

To study the substrate binding of UGT1A5 substrate docking experiments using GOLD [27] were conducted. A representative conformation from the UGT1A5*8 simulation was chosen as protein conformation for docking that shows an opened substrate binding site and optimal geometry of the catalytic center. A common binding mode hypothesis was found for UGT-Glo substrates A and B (Fig. 5). The proposed binding mode includes positioning of the hydroxyl group (UGT-glo substrate A) or primary amine moiety (UGT-glo substrate B) above the anomeric C atom of the glucuronic acid of the cofactor UDP-GA and in close distance to the catalytic His40. Furthermore, hydrogen bonding to Arg196 via the nitrile moiety could be observed for both substrates.

Details are in the caption following the image
Binding mode hypothesis for UGT-Glo substrate A (A) and B (B) to the catalytic site of UGT1A5. The catalytic dyad of His40 and Asp152 (both gray line representation) is in plane with the hydroxyl group of UGT-Glo substrate A and amine group of substrate B (colored green). This allows substrate deprotonation and attack on the anomeric C atom of the glucuronic acid moiety of the cofactor UDP-GA (gray stick representation), which is situated below. Arg174 and Asp397 hold the glucorosyl moiety in place. The flexible Arg196 was rotated to allow for hydrogen bonding to the nitrile moiety in both substrates.

Conclusion

In this study, we characterized two new polymorphic UGT1A5 variants present in the general population, which are a common ninefold mutant (UGT1A5*8) and a much rarer sixfold mutant (UGT1A5*9). UGT1A5*9 displays the same activity as the wild-type, while the activity of UGT1A5*8 is significantly higher for both the O- and N-glucuronidation reactions tested. Molecular modeling studies indicate that the minor effects of all mutations except for Gly259Arg are due to their distance to the substrate binding site. Remarkably, extensive molecular dynamics simulations of UGT1A5*1 and UGT1A5*8 demonstrate that the Gly259Arg mutation stabilizes helix Q through a newly formed hydrogen bonding network, which in turn places the cofactor in a much more favorable geometry in UGT1A5*8 as compared to the wild-type. Thus, we could mechanistically explain how the Gly259Arg mutation increases enzymatic activity. Moreover, our findings show that UGT1A5 does indeed catalyze an N-glucuronidation. Thus, future studies aimed at the formation of N-glucuronides will have to take the activity of UGT1A5 into account.

Author contributions

Participated in research design: GW, MB, Conducted experiments: FY, DM, SW, ZL, Performed data analysis: FY, DM, SW, ZL, GW, MB, wrote or contributed to the writing of the manuscript: FY, DM, GW, MB.